Developing reliable and fast simulated annealing for stand-level forest harvesting schedule with virtual dimensionality reduction

Abstract

Identifying optimal harvesting (thinning and clearcutting) schedule for a forest stand (stand-level) is of great importance to forest managers and researchers. However, providing highly reliable optimal harvesting schedules still requires considerable computation time. This study aimed to develop a simulated-annealing-based method to provide a highly reliable optimal schedule for harvesting in a short time. In practice, the optimal number of harvesting events in a rotation may not be large. Therefore, 1) treating harvesting ages as well as the thinning intensity as control variables, and 2) searching the optimal number of thinning events iteratively, the dimension of the scheduling problem was virtually reduced. The method was developed through a comparison with a reliable method for a model with fixed harvesting ages. Three candidate neighborhood methods were developed and compared with each other. One neighborhood method was less robust than the other two methods. We further developed a method to optimize the planting density simultaneously with the harvesting schedule. Introduction of planting density as a control variable reduced computation time by decreasing the optimal number of harvesting for several cases. The developed method provided optimal schedules with much less time than the previous method for most cases.